There are five different versions of string theory: Type I, Type IIA, Type IIB, Heterotic O(32) (Heterotic-O, for short), and Heterotic E X E (Heterotic-E, for short). Each differs by the way in which it incorporates symmetry and by how strings in the theory vibrate. Since 1995, evidence has mounted indicating that these five theories are actually different ways of looking at the same theory. By adding yet another spacial dimension, the five can be merged into M-theory. M doesn't stand for anything exactly, but perhaps "magic," "mystery," "membrane," "matrix," or the "mother of all theories."

Before 1995, physicists used perturbation theory with string theory. In essence, it is a method of approximating something and then refining the approximation. Physicists must do this because the uncertainty principle dictates that at any time, energy can be temporarily "borrowed" to create string-antistring pairs which then annihilate themselves to "repay" the borrowed energy. The presence of these virtual string pairs, though ephemeral, is enough to effect calculations involving string interactions. The more virtual string pairs erupt into existence, the more calculations are disturbed. Physicists can make approximate calculations by first assuming that no virtual string pairs pop into existence, then increasingly refine their calculations by taking more virtual pairs into account.

The probability that virtual string pairs will appear is called the string coupling constant. It is different for each of the five string theories, none of which is known. A coupling constant less than one means a decreasing likelihood that virtual string pairs will pop into existence, while a coupling constant greater than one represents an increasing likelihood of this occurrence. The catch is that perturbation theory is only valid for coupling constants up to one. So, until 1995, predictions about particles in a universe with a given coupling constant were only made for values up to one.

Then at the Strings ‘95 conference, Edward Witten announced that BPS states could be used to predict particle values with coupling constants greater than one. BPS states predict the values for particles given a chosen charge by utilizing supersymmetry and finding the combination consistent with the given charge while having the minimum mass possible. Using BPS states, Witten found that high values of the coupling constant for Type I (strongly-coupled Type I) string theory match the properties of weakly-coupled (less than one) Heterotic-O string theory. They are dual, meaning they seem different but create the same physical results. This kind of relationship is called strong-weak duality. It was soon discovered that Type IIB string theory is self-dual, meaning that reciprocal values for its coupling constant yield matching results. This glimpse of interrelation led physicists to believe that they were onto something big - the unification of the five kinds of string theory.